Update model card

README.md

@@ -12,24 +12,22 @@ tags:
 - apple-amlr
 ---
 
-# ml-sharp-webgpu
+# Sharp-webgpu
 
-This repository contains a browser-oriented ONNX/WebGPU export of Apple's SHARP
-monocular view-synthesis model. It is a derived artifact of the official
-[`apple/Sharp`](https://huggingface.co/apple/Sharp) checkpoint, packaged for the
-`sharp_webgpu.html` browser runtime in
-[`apple/ml-sharp`](https://github.com/apple/ml-sharp).
+This repository contains a staged ONNX export of Apple's SHARP monocular
+view-synthesis model. It is a derived artifact of the official
+[`apple/Sharp`](https://huggingface.co/apple/Sharp) checkpoint.
 
 This is not the official Apple checkpoint repository and is not endorsed by
-Apple. The goal of this repo is to make the released research model easier to
-run in WebGPU-capable browsers through the SHARP WebGPU pipeline and
-Transformers.js v4.
+Apple. The package contains model graph files, external tensor data, checksums,
+licenses, and a manifest describing the staged artifact layout.
 
 ## Intended Use
 
 - Non-commercial scientific research and academic development.
-- Browser-side WebGPU inference for SHARP image-to-splat and panorama workflows.
-- Local validation of the staged ONNX export against the SHARP WebGPU runtime.
+- Evaluation of a staged ONNX derivative of the released SHARP research model.
+- Integration into research clients that implement the required SHARP
+  preprocessing, staged inference, and postprocessing steps.
 
 ## Not Intended For
 
@@ -48,7 +46,7 @@ That license limits use, modification, derivatives, and redistribution to
 research purposes and excludes commercial exploitation, product development, or
 commercial product/service use.
 
-This ONNX/WebGPU bundle is a derived artifact from the released SHARP model and
+This ONNX artifact bundle is a derived artifact from the released SHARP model and
 should be treated as a model derivative for licensing purposes.
 
 Attribution required by `LICENSE_MODEL`:
@@ -56,65 +54,10 @@ Attribution required by `LICENSE_MODEL`:
 > Apple Machine Learning Research Model is licensed under the Apple Machine
 > Learning Research Model License Agreement.
 
-Modifications from the upstream `apple/Sharp` release:
-
-- Exported the PyTorch checkpoint to a staged ONNX layout for browser execution.
-- Split execution into `stage-1` through `stage-4` for ONNX Runtime WebGPU.
-- Applied the ONNX Runtime 1.26 compatibility export/patch that removes
-  `Cast(to=DOUBLE) -> IsInf` adapters emitted by PyTorch `nan_to_num`.
-- Added a `manifest.json` consumed by the SHARP WebGPU runtime.
-
-## Files
-
-```text
-README.md
-LICENSE
-LICENSE_MODEL
-.gitattributes
-config.json
-manifest.json
-sharp_stage_1.onnx
-sharp_stage_1.onnx.data
-sharp_stage_2.onnx
-sharp_stage_3.onnx
-sharp_stage_4.onnx
-SHA256SUMS
-```
-
-## Browser Usage
-
-Serve the SHARP WebGPU app and load this repository through Transformers.js:
-
-```text
-http://localhost:8000/sharp_webgpu.html?model=ml-sharp-webgpu
-```
-
-For local validation before upload, place this folder under `webgpu/models/` and
-open:
-
-```text
-http://localhost:8000/sharp_webgpu.html?model=ml-sharp-webgpu&local=1&localModelPath=./models/
-```
-
-## Validation
-
-This folder was prepared with:
-
-```bash
-./.venv/bin/python webgpu/prepare_hf_release.py --repo-id ml-sharp-webgpu
-```
-
-Before publishing, run:
-
-```bash
-./.venv/bin/python webgpu/prepare_hf_release.py \
-  --repo-id ml-sharp-webgpu \
-  --validate-only
-```
+Artifact modifications from the upstream `apple/Sharp` release:
 
-The validation checks that all expected files are present, external ONNX data is
-available, SHA-256 hashes match, license markers exist in this model card, and
-`sharp_stage_1.onnx` no longer contains the ORT 1.26 double-cast export pattern.
+- Exported the PyTorch checkpoint to a staged ONNX layout.
+- Split execution into `stage-1` through `stage-4`.
 
 ## Citation